Research

Explainable Artificial Intelligence (AI) and Benefits for Business Applications


The National Institute of Standards and Technology (NIST) recognized the many challenges in designing, constructing and assuring a Cyber Physical System (CPS), and in response developed the CPS Framework, designed to break down the process of completing a CPS into three separate facets and to aid in the processes associated with the realization of a CPS. We augment the Framework with AI and Explainable AI (XAI), through which we provide the tools promoting cross-functional collaboration and offer three high-level approaches to understand and explain the decisions made by AI that will be explored throughout the poster.

https://sites.google.com/sju.edu/nedsi21poster/home


Poster Presentation 2021 NEDSI Annual Conference, Harrisburg, PA.

Intelligent Agents and Cognitive Robotics

We conduct research on integrating knowledge representation and reasoning techniques with robotics. Our goal is to endow robots with sophisticated, human-like behavior. Our work includes the design and implementation of logic formalisms, representation methodologies, and logic-based reasoning algorithms. We focus on domains that are knowledge intensive and require sophisticated reasoning capabilities. Ultimately, we aim to define agent architectures capable of planning, diagnostics, and execution monitoring in the kinds of domains a realistic robot might encounter and under the computational constraints a realistic robotic architecture would impose. Our projects include:

  • Languages, solvers and reasoning mechanisms for hybrid qualitative-quantitative knowledge

  • Planning, diagnostics, learning in discrete and hybrid discrete-continuous domains

  • Agent architectures

Natural Language Understanding & Reasoning about Actions and Change

We believe that sophistication in Natural Language Understanding can only be achieved if agents are endowed with the ability to understand how the world described by natural language passages evolves as the story progresses. Doing so requires being able to reason about the direct and indirect effects of actions, about the plausible causes of observations, about goals and intentions of agents, and to use such reasoning processes to identify ways of filling the gaps in a story.

To achieve all this, we conduct research on the integration of Natural Language Understanding and Reasoning about Actions and Change. Our projects include:

  • Causality and narratives

  • Action recognition in narratives

  • Information retrieval in the presence of actions and change

  • Theories of intentions

Cyber Security

Our work in cybersecurity is aimed at increasing the breadth of understanding of threats and of mitigation strategies by integrating cybersecurity, knowledge representation and automated reasoning. Our projects include:

  • High-level querying of log files and other sources with unknown of partially known format

  • Automated mitigation of threats

  • Study of roots of trust in cyber-physical systems

  • Study of issues, attack vectors, defense mechanisms, analysis techniques in automotive security